If I want to adapt a foundational base model(i.e. Llama or GPT-J-6B etc.) for a custom domain and have many unstructured documents, how to feed these documents into a foundational base model? I guess I need to do this step before instruction fine tuning and RLHF so that the model has the knowledge regarding the custom domain first. Is this correct?
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
Adding domain knowledge in LLMs via fine tuning | 2 | 5447 | July 23, 2023 | |
Primer on Fine Tuning Text generation models (like GPT) | 0 | 1373 | November 14, 2022 | |
Finetuning llm model using Trainer and reusing it | 3 | 1206 | August 17, 2023 | |
Base or Instruct version of LLM for fine tuning? | 1 | 2634 | August 8, 2024 | |
Fine-tuning conversational models with the technical documentation | 2 | 1258 | July 18, 2024 |